"Two-thirds of those released from prison in the United States will be re-arrested within three years, creating an incarceration cycle that is detrimental to individuals, families, and communities." So writes Jennifer L. Doleac in "Strategies to productively reincorporate the formerly-incarcerated into communities: A review of the literature" (posted on SSRN, July 21, 2018), Doleac's approach is straightforward: look at the studies. In particular, look at fairly recent studies done since 2010 that use a "randomized controlled trial" approach--that is, an approach where a group of participants are randomly assigned either to receive a particular program or not to receive it. When this approach is carried out effectively, comparing the "treatment group" and the "control group" provides a reasonable basis for drawing inferences about what works and what doesn't.

Here's a list of the interventions on which Doleac finds some fairly recent studies using randomized controlled trial approaches. Some of the studies focus on recidivism, while others look at outcomes like employment or gaining additional education. 

I'll let you read Doleac's literature review for details of individual studies. But I'll just notice here that this kind of list does not seek to respect what one expects or hopes might be true. 

For example, the "bad bets" at the bottom all have their advocates. But based on the evidence, Doleac writes concerning these programs: 
"Many programs focus on increasing employment for people with criminal records, with the hope that access to a steady job will prevent reoffending. This topic has been studied more than others, and the research results are mixed. Transitional jobs programs provide temporary, subsidized jobs and soft-skills training to those trying to transition into the private sector workforce. multiple rigorous studies show that transitional jobs programs are ineffective at increasing post-program employment, and have little to no effect on recidivism. ...

"Ban the Box policies seek to increase access to employment by prohibiting employers from asking about criminal records until late in the hiring process. Research shows that Ban the Box policies are ineffective at increasing employment for people with criminal records,and have the unintended consequence of reducing employment for young black men withoutcriminal records (because employers assume that applicants from this group are more likely to have a record when they cannot ask directly). The net effect is a reduction in employment for young, low-skilled black men--the opposite of what proponents of this policy hoped to achieve. ... 
"Given the array of challenges faced by people who cycle through the criminal justice system, a popular approach is to try to address many needs at once. Two evaluations of highly-respected reentry programs providing wrap-around services found little to no effect on subsequent recidivism. More recently, two large-scale evaluations of federal programs funding wrap-around services in communities across the country both found increases in recidivism for the treatment groups. ... Together, these studies suggest that these multi-faceted, labor-intensive (and thus expensive) interventions may be trying to do too much and therefore do not do anything well. Since this is a popular approach in cities and counties across the country, leaders should be skeptical about the effectiveness of their current programs."
Conversely, here are some comments on what seems most promising, based on the actual studies. From Doleac:
"Court-issued rehabilitation certi ficates can be presented to employers as a signal of recipients' rehabilitation. One study found that court-issued certi ficates increased access to employment for individuals with felony convictions. This could be because they provide valuable information to employers about work-readiness, or because employers perceive the court-issued certi ficates as protection against negligent hiring lawsuits. In either case, this strategy is promising and worth further study. The effect on recidivism is currently unknown. ...
"A large share of people who are arrested and incarcerated suffer from mental illness, and many more are hindered by emotional trauma and poor decision-making strategies. Therapy and counseling could have a meaningful impact on the successful reintegration of these individuals. Programs focused on mental health include cognitive behavioral therapy (CBT) and multisystemic therapy (MST). A growing body of evidence supports CBT as a cost-effective intervention, though the evidence on MST is more mixed and may be context-dependent. In both cases, it is unclear how much effectiveness will fall if programs are scaled up to serve more people: if they require highly-trained psychologists to conduct the sessions, the scalability will be limited. ...

"Diverting low-risk offenders to community supervision instead of incarceration appears to be highly effective. Electronic monitoring is used as an alternative to short incarceration spells in several countries, and in those contexts has reduced recidivism rates and increased economic well-being and educational attainment. Court deferrals--which allow low-risk, non-violent felony defendants to avoid a conviction if they successfully complete probation--reduce recidivism rates and increase employment. And an innovative diversion program for non-violent juvenile offenders that provides group mentoring and instruction in virtue theory was shown to reduce recidivism relative to standard diversion to community service. ...
"Many people coming out of jail or prison may benefit from government or community support, but many others might be better off if we left them alone. (This is especially likely if the programs they would be referred to are not effective.) A diverse set of high-quality studies consider the effects of reducing the intensity of community supervision. All found that reducing intensity of supervision (for example, requiring fewer meetings or check-ins with probation officers) has no impact on recidivism rates, and that it actually reduces recidivism for low-risk boys (age 15 or younger). That is, for less money, and less hassle to those who are court-supervised, we could achieve the same and even better public safety outcomes. This approach is worth exploring in a variety of contexts, and appears to be effective for high-risk as well as low-risk offenders. ... At this point, there is substantial evidence, from a variety of contexts, that increasing the intensity of community supervision has no public safety benefi ts and in some cases increases recidivism. It is also more expensive. It is unclear what the optimal amount of supervision is for various types of offenders, but it's clearly lower than current levels. ... 
"[A]nother policy that has great potential to reduce recidivism and incarceration
rates is expanding DNA databases. Two studies show that those charged or convicted of
felonies are dramatically less likely to reoffend when they are added to a government DNA database, due to the higher likelihood that they would get caught. Deterring recidivism in this way is extremely cost-effective, and reveals that many offenders do not need additional supports to stay out of trouble."
Doleac emphasizes that the evidence on many of these programs is not as strong as one might prefer, and there is certainly room for more research. But I would add that those looking to go beyond research and enact a wide-ranging alteration of policies should be considering the existing research, too.